🚀 Crypto-Momentum Dashboard & Chatbot (v1.0 Final)

📌 Project Overview

The Crypto-Momentum Dashboard & Chatbot project is an integrated financial AI platform designed for cryptocurrency price forecasting, primarily focusing on Bitcoin (BTC) and Ethereum (ETH). It leverages historical price data and market sentiment from news and social media sources to provide real-time, actionable insights through Machine Learning (ML) and Deep Learning (DL) techniques.

Important Notices

• ✅ Price Data: Continuously updated via automated pipelines (yfinance)
• ⚠️ Sentiment Data: Static data (last update: May 25, 2025), limited due to external sources
• ⚠️ NLP & LLM Chatbot Modules: Deferred to future phases

🎯 Objectives & Capabilities

✅ Completed Features

Data Engineering & Preprocessing:

• Daily OHLCV data (2018–2025) from Yahoo Finance
• Aggregated daily sentiment scores (limited to May 2025)
• Feature engineering pipeline (lag features, technical indicators)

Exploratory Data Analysis (EDA):

• Visualization of price-sentiment relationships
• Identification of high-impact news periods

Machine Learning (ML):

• LightGBM Classifier: Price direction (binary classification for +2% moves over 1, 3, and 5 days)
• Walk-forward validation (realistic backtests)

Deep Learning (DL):

• Single-task & Multi-task LSTM: Capturing short/medium-term trends
• 1D-CNN Architecture: Experimental study (not deployed due to lower performance)

Explainability:

• SHAP analysis for feature importance (summary/waterfall plots)

Strategy & Backtesting:

• Capital allocation simulations (theoretical returns vs. buy-and-hold)

Interactive Dashboard:

• Real-time ML & DL predictions via Streamlit
• Visual analytics (heatmaps, confusion matrices, equity curves)

🛠️ Future Enhancements

• Real-time NLP sentiment analysis
• LLM-powered conversational chatbot (FastAPI, Azure OpenAI / Hugging Face)

📊 Technical Details & Architecture

Layer	Technologies & Libraries
Data Engineering	Python, Pandas, NumPy, yfinance, Parquet
ML Modeling	scikit-learn, LightGBM, joblib
DL Modeling	TensorFlow, Keras
Explainability	SHAP
Dashboard	Streamlit, Plotly
Chatbot (Planned)	FastAPI, Azure OpenAI / Hugging Face
Storage	Parquet, CSV
Version Control	Git, GitHub

📁 Folder Structure

Project/
├── data/
│   ├── raw/                  # Raw BTC/ETH/news data files
│   ├── processed/            # Feature-engineered data
│   └── dashboard_data/       # Streamlit-ready data
│
├── src/                      # Data pipeline & ML scripts
│   ├── 01_data_processing.py
│   ├── 02_merge_data.py
│   ├── 03_data_validation.py
│   ├── 04_clean_merged.py
│   ├── 05_feature_engineering.py
│   ├── 06_train_walkforward.py
│   ├── 07_detailed_report.py
│   ├── 08_threshold_tune.py
│   ├── 09_prepare_dashboard_data.py
│   ├── 11_shap_values.py
│   └── backtest_helper.py
│
├── dl/                       # Deep Learning workflows
│   ├── seq_window.py         # Sliding window data prep
│   ├── train.py              # Training LSTM & CNN
│   ├── lstm.py               # Single-task LSTM
│   ├── lstm_multitask.py     # Multi-task LSTM
│   ├── cnn.py                # CNN (experimental)
│   ├── evaluate.py           # Evaluation scripts
│   ├── outputs/              # Prepared datasets (.npz)
│   └── scalers/              # Scaler objects (.pkl)
│
├── models/                   # ML (.pkl) & DL (.h5) models
├── results/                  # Reports & visualizations
├── notebooks/                # Jupyter notebooks (EDA)
├── streamlit_app.py          # Streamlit dashboard
└── requirements.txt          # Dependencies

🚀 Installation & Usage

Step 1: Clone Repository

git clone https://github.com/Taylanozveren/Ai_Project_Chatbot_Fintech.git
cd Ai_Project_Chatbot_Fintech

Step 2: Virtual Environment Setup

python -m venv venv
source venv/bin/activate        # Windows: venv\Scripts\activate
pip install -r requirements.txt

Step 3: Run Data Pipeline & ML Scripts

python src/01_data_processing.py
python src/02_merge_data.py
python src/03_data_validation.py
python src/04_clean_merged.py
python src/05_feature_engineering.py
python src/06_train_walkforward.py
python src/07_detailed_report.py
python src/08_threshold_tune.py
python src/09_prepare_dashboard_data.py
python src/11_shap_values.py

Step 4: Train & Evaluate DL Models

python dl/seq_window.py
python dl/train.py
python dl/evaluate.py

Step 5: Launch Streamlit Dashboard

streamlit run streamlit_app1.py

⚠️ Current Limitations & Notices

• NLP Module: Static sentiment data up to May 25, 2025 (due to external source constraints)
• Chatbot & LLM Integration: Planned future enhancements (pending development)

📅 Roadmap & Next Steps

Timeline	Milestone	Status
June 2025	ML & DL Dashboard Finalized ✅	Completed
June 2025	Deployment of Dashboard ✅	Completed
July 2025+	Real-time NLP Integration (Currently Just Invalid csv Data 🛠️	Planned
July 2025+	LLM-Powered Chatbot 🤖	Planned

🤝 Contribution & Collaboration

Lead Developer & Architect: Taylan Özveren
📧 Contact: [email protected]

Contributions are highly encouraged:

1. Fork the repository
2. Create your branch:

   git checkout -b feature/MyNewFeature

3. Commit changes:

   git commit -m "Add MyNewFeature"

4. Push to branch:

   git push origin feature/MyNewFeature

5. Open Pull Request on GitHub

📃 License

This project is licensed under the MIT License - see LICENSE.md for details.